One endpoint. The best model, every call.
A routing model Actualyze trains sends every request to the cheapest model strong enough for it, and stays current as models change. The same model serves through any provider you already use, with standby routes ready when a route fails.
Composable base models with smart routing built in.
A virtual model is a stable address in front of several distinct models, like a virtual IP for inference. Route across them by the capabilities and cost-quality preference you want, so simpler requests can go to a cheaper or faster model.
- Pick the models behind the virtual model
- Define route preferences by cost, quality, and capability
- Preview the route before saving, then go live the moment you save
- Swap underlying models without code changes
- Routing never widens access: a call reaches only models the caller may already use
Route to the best model for every call.
Point your app at a virtual model; Actualyze picks a real model for each request, using the cost-quality preference you set and the capabilities you require. Model Groups cover the same model across providers: a Workspace Admin declares that per-provider catalog rows are one model, so a call can be served through any provider the caller may already use. You get resilience across providers without loosening permissions.
- Capability-based routing across distinct models
- Use virtual models to optimize spend, get the best quality response, or both
- Model Groups serve one model through any provider the caller already uses, with no new permissions
- Automatic failover down an ordered list when a model errors or stalls
- Each request sent to the cheapest model strong enough for it
See every route, in real time.
A route is one model through one endpoint, and each route carries its own guardrails and price. That price is the rate you confirmed, else the seller's list rate, else the catalog reference; a cost nobody can price shows as unknown, never as $0. Budget enforcement and your data rules apply on every routed call.
- Every audit record names the model asked for and the route that served it
- Latency and cost visibility per virtual model
- See the models used and why they were chosen
- Get request volume, routes, and costs by team or model
- A routing mix of standard, upgraded, cost-save, and failover calls
Explore the rest of the platform.
Every call, automatically optimized.
Stop integrating multiple providers into your code. Give your apps one endpoint and let the platform do the rest.